Research On Power System Prediction Task Based On Improved Temporal Convolution Network

With the extensive research and application of deep learning algorithm represented by neural network,power systems all over the world also actively introduce deep learning algorithm to improve the prediction accuracy of traditional prediction tasks.This paper mainly studies the load forecasting of power system and the reheater wall temperature forecasting of thermal power plant.This paper analyzes the data characteristics of the two prediction tasks in detail.In the research of load forecasting and reheater temperature forecasting,the algorithm structure of temporal convolution network is improved to construct an accurate prediction model separately,and high-precision prediction results are obtained.Aiming at the task of power system load forecasting,this paper mainly carries out the following analysis and experiments.With the advancement of power market reform,accurate load forecasting can ensure the stable operation of power systems increasingly.The randomness of feature change such as climate increases the complexity of short-term load forecasting.To predict short-term loads more accurately,this paper takes the past load data as a feature and considers the time series characteristics of load data simultaneously.The multi-temporalspatial-scale method is applied to process the load data by reducing the load data noise error and enhancing the time series characteristics.Then,a novel short-term load forecasting model,which is named a multi-temporal-spatial-scale temporal convolutional network,is applied to load forecasting tasks in this paper.The proposed approach can learn the nonlinear feature and time series characteristics of load data simultaneously.To predict the power load of a city in the next day and the next week,the forecasting model is trained by the historical feature load of 7 days,21 days,99 days,and 199 days.Compared with 22 traditional shortterm load forecasting models,such as artificial neural network,the simulation results show that the proposed multi-temporal-spatial-scale temporal convolutional network can obtain higher accuracy for the short-term load forecasting of power systems than other compared methods.Aiming at the task of reheater wall temperature forecasting in thermal power plant,the following analysis and experiments are carried out in this paper.The ultra-supercritical boiler has higher reheater steam temperatures.The possibility of overtemperature tube explosion of high temperature reheater increases.Therefore,it is necessary to predict the distribution state of high temperature reheater wall temperature and set alarm threshold to prevent the occurrence of failures such as tube explosion.The multi-feature-scale fusion temporal convolution network prediction model is proposed in this paper.The model has a strong ability of feature extraction and nonlinear function fitting.The multifeature-scale fusion method can fully consider the time delay between the input material of the coal-fired boiler system and the wall temperature change of the high temperature reheater.This method is the key technology to accurately predict the wall temperature of high temperature reheater.Original data are collected from one thermal power plant reheater.The multi-feature-scale fusion temporal convolution network is applied to study the real-life data from March 2020 to May2020.Compared with 20 prediction models such as artificial neural network,the proposed model has higher accuracy.